Deep Learning-Based Virtual Elastin Staining Improves Visceral Pleural Invasion Assessment in Lung Cancer.

Area of Science:

• Digital Pathology
• Artificial Intelligence in Medicine
• Oncology Diagnostics

Background:

• Accurate assessment of visceral pleural invasion (VPI) is crucial for non-small cell lung cancer (NSCLC) staging and prognosis.
• Distinguishing elastin-rich pleural layers on routine H&E slides is diagnostically challenging.
• Special elastic stains are costly and cause workflow delays.

Purpose of the Study:

• To develop a deep learning pipeline for virtual elastin staining (synthetic EBEF) from H&E slides.
• To overcome limitations of traditional elastic stains in VPI assessment.
• To improve diagnostic accuracy for VPI in NSCLC.

Main Methods:

• A conditional generative adversarial network was trained using intrinsic eosin fluorescence as ground truth.
• Synthetic eosin-based elastin fluorescence (EBEF) was generated from standard brightfield H&E slides.
• Multi-institutional validation assessed the impact of synthetic EBEF on VPI diagnostic accuracy.

Main Results:

• Synthetic EBEF significantly improved pathologists' diagnostic accuracy for VPI compared to H&E alone (P < 0.0001).
• Thinner tissue sections (1-3 μm) and high-resolution scanning optimized model performance and elastin contrast.
• A strong synergy was observed between computational and conventional diagnostic optimization.

Conclusions:

• A robust framework for high-fidelity virtual staining was established and validated.
• The synthetic EBEF approach offers a practical, cost-effective alternative to special stains for NSCLC evaluation.
• This deep learning tool provides a scalable pathway for integrating AI into routine digital pathology.